A Water-Promoted Mars−van Krevelen Reaction Dominates Low-Temperature CO Oxidation over Au-Fe2O3 but Not over Au-TiO2

We provide experimental evidence that is inconsistent with often proposed Langmuir−Hinshelwood (LH) mechanistic hypotheses for water-promoted CO oxidation over Au–Fe2O3. Passing CO and H2O, but no O2, over Au-γ-Fe2O3 at 25 °C, we observe significant CO2 production, inconsistent with LH mechanistic hypotheses. Experiments with H218O further show that previous LH mechanistic proposals cannot account for water-promoted CO oxidation over Au-γ-Fe2O3. Guided by density functional theory, we instead postulate a water-promoted Mars–van Krevelen (w-MvK) reaction. Our proposed w-MvK mechanism is consistent both with observed CO2 production in the absence of O2 and with CO oxidation in the presence of H218O and 16O2. In contrast, for Au-TiO2, our data is consistent with previous LH mechanistic hypotheses.


S1.2. Cleaning procedures
A base bath (8 L of isopropyl alcohol, 2 L of deionized (DI) water, and 500 g of KOH) was used to clean all glassware.After immersion for at least 1 h in the base bath, the glassware was rinsed copiously in DI water and then in Milli-Q water, followed by drying in a clean oven at 80 o C.

S1.3.Preparation of catalysts
We adapted a well-established deposition-precipitation protocol from the literature 1 for synthesis of catalysts with 1 % weight loading of Au.Briefly, 0.06 g of HAuCl4 • 3 H2O was added to a 500 mL reaction flask, using a glass-boat and a Teflon spatula.300 mL water was then added, and the HAuCl4 • 3 H2O allowed to dissolve.Urea (8 g) was then added to the reaction flask, and allowed to dissolve, followed by addition of 3 g of the support (γ-Fe2O3 or TiO2) .The solution was then sonicated (Elmasonic P, Elma) for 2 min.The reaction flask was capped with septa, including a 23 gauge needle for venting, and a glass-thermometer inserted through the septa.The solution was then stirred with a magnetic stirbar (700 rpm) and heated to 80 o C +/-3 o C in an oil-bath.The reaction was then stirred for 16 hours.The reaction was then brought to 50 o C on a water-bath.The suspension was then transferred to falcon tubes and the product washed repeatedly by centrifugation (Heraeus Megafuge 16, Thermo Scientific) followed by re-suspension in water at 50 o C: -Centrifugation at 10000 rpm (3 min), discarding the supernatant followed by re-suspending (sonication) the sample in water at 50 o C for 10 min.
-Centrifugation at 10000 rpm (3 min), discarding the supernatant followed by re-suspending (sonication) the sample in water at 50 o C for 10 min.
-Centrifugation at 10000 rpm (3 min), discarding the supernatant followed by re-suspending (sonication) the sample in water at 50 o C for 10 min.
-Centrifugation at 10000 rpm (3 min), discarding the supernatant followed by re-suspending the sample in water at 50 o C for 10 min.
-Centrifugation at 10000 rpm (3 min), collecting the sample and then drying the sample in desiccator for 36 h.Finally, the sample was annealed in air at 250 o C for 4 hours to produce the catalysts (Au-γ-Fe2O3 and Au-TiO2, respectively, 1 wt % Au.).At all times during sample preparation and storage, the samples were protected from light.BET and XRD analysis of the catalysts are presented in Figure S10.

S1.4. Characterization Techniques
Transmission electron microscopy (TEM) images were collected on a JEOL JEM-2100F instrument operating at 200 kV.The samples were deposited onto lacey carbon / Cu grids by simply shaking the grid in a small amount of sample powder.High-resolution TEM images by aberration-corrected TEM (Thermo Fisher Themis Z) operated at 300 kV.Spherical aberration was compensated using an image corrector (CEOS CETCOR) up to third order.TEM images were acquired by Gatan OneView camera.
Nitrogen adsorption isotherms for Brunauer-Emmett-Teller (BET) analysis were recorded on a Micromeritics ASAP2020 analyzer at 77 K. Samples were degassed at 100 °C for 12 hours prior to measurement.Powder X-ray diffraction patterns were recorded on a Bruker D8 DISCOVER diffractometer applying Cu Kα radiation, with a scanning speed of 4° min −1 in a 2θ range of 10° -80° in reflection mode.

S1.5. Catalytic tests
CO oxidation rates were measured in a packed-bed plug flow reactor (quartz tube).The catalyst bed (40 mg catalyst) was placed between 2 layers of acid-washed sand (200 mg bottom layer, 280 mg top layer).
Without catalyst, the sand was not active for CO oxidation under the reaction conditions used.All reactions were carried out at ambient pressure, and the reaction temperature was controlled at 25 o C. Gas flows were obtained by mixing CO with either a mixture of 1 vol % Kr in N2 or with a mixture of 20 vol % O2 and 1 vol % Kr in N2, where Kr was used as an inert internal standard.Gases were mixed with electronic mass-flow controllers (EL-FLOW prestige, Bronkhorst).Water was introduced in the gas stream by a bubbler (15 mL H2O) kept at 23 o C, corresponding to a concentration of 2.8 vol % H2O.
Concentrations in the reactor effluent were measured, in-line, by mass-spectrometry (CIS 300, SRS), using 1 vol % Kr as an internal standard (all m/z signals were normalized to the m/z = 86 signal where the m/z = 86 signal corresponds to 86 Kr).
Before measurements, ambient gases adsorbed to the catalyst were removed by passing humidified N2 (2.8 vol % H2 16 O or H2 18 O) over the catalyst for 16 h.The reaction mixture was then introduced; {1 vol % CO, 1 vol % Kr, 2.8 vol % H2O, balance N2} for measurement in absence of O2, and {1 vol % CO, 1 vol % Kr, 20 vol % O2, 2.8 vol % H2O, balance N2} for measurement in presence of O2.The flow rate was 14 mL min -1 , corresponding to a gas hourly space velocity (GHSV) of 21 L h - 1 g -1 cat.These conditions led to CO conversions below 20 %, thereby ensuring differential conditions, allowing direct comparison of measured rates. 2,3For C 16 O oxidation experiments using 16 O2 and H2 18 O, the same conditions as described above were used, with the difference that H2 16 O was replaced by H2 18 O, and a flow-rate of 7 mL min -1 (instead of 14 mL min -1 ) was used in the H2 18 O experiments to allow prolonged bubbling through the small volume (1 mL) of H2 18 O, resulting in a GHSV of 10.5 L h - 1 g -1 cat.

S2.1. Computational details
Periodic DFT calculations were performed in Vienna Ab-initio Simulation Package (VASP), 4 employing the PBE 5 exchange-correlation functional.Standard PBE PAW potentials 6 were used to represent core states while the valence states were treated explicit by plane-wave basis-set with an energy cutoff of 520 eV.Γ-centered Monkhorst-Pack type k-points meshes using tetrahedron method and Blöchl corrections. 7The first Brillouin zone was sampled on (3×3×1) for slab and (1×1×1) for molecules.First-order Methfessel-Paxton 8 scheme with a Gaussian width of 0.15 eV was employed for all structures.Structure relaxations were conducted within the criteria of 10 -5 eV and 0.015 eV/Å for energy and atom force, respectively.From the optimized bulk structure, γ-Fe2O3 and Au surface models were constructed in the atomic simulation environment (ASE). 9For all slab structures a vacuum space of more than 20 Å was used, and dipole correction was used along slab z-direction. 10Spin-polarized calculations were carried out, and the most stable spin state for all systems was used in this work.
Molecules were optimized in a cubic box of 20 Å. Transition states (TS) were obtained by CI-NEB (Climbing Image -Nudged Elastic Band) method. 11Nine images were used, and each image converged to the minimum energy pathway using a convergency accuracy of 0.05 eV/Å.The activation energy (  ) was obtained by the difference between the initial state (IS) and TS total energy,   =   −   .
The unit cell used for bulk calculations was obtained by Pecharroman et al. 12 crystallographic refinement data.Maghemite (γ-Fe2O3) crystalizes at room temperature in a cubic system and belongs to Fd3m space group.The experimental cell parameters were found to be {a=b=c= 8.35 Å, α=β=γ= 90.0°}, while our bulk optimizations found the cell parameters {a=b=c= 8.48 Å, α=β=γ= 90.0°}, within 2% of experimental values.This result clearly shows that DFT/PBE level can describe crystal structure of maghemite.
The lattice parameters of Au (A) and γ-Fe2O3 (B) phases cannot be perfectly matched.Thus, it is necessary to evaluate the commensurability of the phases.The natural stress generated by the two phases in contact (A/B) is typically measured by the mismatch parameter (ξ).
Here   and   are the surface area, and  / is the overlap surface area.A detailed description of how to obtain the surface area values has been previously reported. 13,14As illustrated in Table S1, Au(111) and γ-Fe2O3(111) surfaces comprises the lower value for mismatch displacement (~3.4 %).Because of that Au(111)/γ-Fe2O3(111) was selected as a representative interface for the calculations.To compute the reaction energies at the interface formed between γ-Fe2O3 substrate and Au   S2.

S2.2 Free-energy calculations
Reaction free energies (∆G) were calculated based on the equation below: where  +/− 2  3 and  /− 2  3 are the energies of the surface with and without the adsorbed reaction intermediates, respectively.The sum of reactant/product gas phase molecules (H2O, O2, CO, and CO2) is defined as: The free energy of the molecules (  ) were calculated as where ∆µ  0 ( 0 ,  0 ) represents the chemical potential of the molecules at standard conditions, i.e.,  0 = 1 bar and  0 =298.15K.All the values for ∆µ  0 ( 0 ,  0 ) were obtained in NIST-JANAF tables, 16 and the values for H2O, O2, CO, and CO2 are 0.48, 0.54, 0.53, 0.58 eV, respectively.To combine theoretical calculations with the experiments performed in the present work, the partial pressures of the molecules (  ) were considered equal to 0.028, 0.200, 0.010 and 0.010 bars for H2O, O2, CO, and CO2, respectively.(Scheme 1, main paper and Table S2).Note that some reactions (R3, R5, R7) were divided into several elementary steps, to enable DFT computation.Reaction free energies and activation free energies in eV.
Figure S3.Elementary reactions used for DFT calculation of reaction energetics for Iglesia's proposed LH mechanism 2 (Scheme 2, main paper and Table S2).Note that we allowed this reaction to occur at the Au-γ-Fe2O3 interface, (as opposed to exclusively on the Au NP, as proposed by Iglesia).The reason is that the reaction on a pure gold surface is not feasible, since the adsorption free energy for O2 adsorption on Au(111) was calculated to 0.99 eV (R9, Figure S4 and Table S2).Note that some reactions (R12, R13, R14) were divided into several elementary steps, to enable DFT computation.Reaction free energies and activation free energies in eV.Structures     The gas hourly space velocity (GHSV) was 21 L h -1 g -1 cat, and the CO conversion was below 20 % (ensuring data was collected under differential conditions). 2,3Note that the CO2 rate data is the same as the data presented in Figure 2 (B), main paper.Here the CO2 production is compared with the concomitant H2 production, which is negligible (therefore ruling out that the CO2 production is due to the water-gas shift reaction).Reported curves are averages of three independent measurements.Error bars are 2 standard deviations wide.For some data points, the error-bars are so small, they are obscured by the data-labels.In the presence of 16 O2 and H2 18 O the previously proposed LH-mechanisms and our proposed w-MvK mechanism are expected to lead to different abundances of C 16 O2, C 16 O 18 O and C 18 O2.To calculate the abundances predicted by these mechanisms, first we must consider that all CO2 species (C 16 O2, C 16 O 18 O and C 18 O2) exchange oxygen directly with H2 18 O in our reactor set-up.Then, we must separate this exchange from the 18 O incorporation from H2 18 O into CO2 due to the CO oxidation reaction mechanism.
To estimate the direct 18 O exchange between H2 18 O and different isotopic CO2 species, we first mixed C 16 O2 (0.34 mol %), H2 18 O (2.8 mol %) and 16 O2 (20 mol %), balance N2, and then passed this mixture over the Au-γ-Fe2O3 catalyst, at a flow-rate (8.7 mL min -1 ) which is similar to the flow-rate (7 mL min -1 ) used for C 16 O oxidation with H2 18 O and 16 O2 (Figure 5).The slightly higher flowrate in the control experiment was used to allow the C 16 O2 concentration in the control to correspond to the maximum total CO2 concentration obtained during C 16 O oxidation with H2 18 O and 16 O2.The time evolution of the relative fractions (%) of C 16 O2, C 16 O 18 O and C 18 O2 in the reactor effluent, are presented in Figure S12.At time t = 0 h the flow was turned from bypassing the catalyst bed, to pass the catalyst bed.At this time, the relative abundances are 93.2 % (C 16 O2), 6.3 % C 16 O 18 O and 0.4 % (C 18 O2), suggesting there is some 18 O exchange from H2 18 O into C 16 O2 in the piping.Upon turning the flow over the catalyst bed, new steady state relative ratios are established, namely 57.5 % (C 16 O2), 35.9 % (C 16 O 18 O) and 6.6 % (C 18 O2).These steady-state abundances can be interpreted as a property of the reactor/catalyst-bed set-up, and can be translated into the probability that a C 16  From these equations we can also derive the probability that a C 16 O 18 O molecule transforms into C 18 O2 in the reactor/catalyst-bed set-up.First, we note that a single 18 O exchange leads to C 18 O2 with half the probability given by equation S6 (0.359/2), because in C 16 O 18 O half of the exchanges are with the 16 O and other half with the 18 O.Then, we observe that double 18 O exchange in C 16 O 18 O, which also produces C 18 O2, should have the same probability as the 18 O exchange with both 16 O in C 16 O2, (0.066, equation S7).Therefore, the probability of forming C 18 O2 by 18 O exchange with C 16 O 18 O is given by the combined probability of single and double 18  where we reproduce Figures S2 -Figure S5, but using H2 18 O instead of H2 16 O.
We now have knowledge of 18 O exchange probabilities between different isotopic CO2 species and H2 18 O in our reactor (equations S6 -S8) and we have predicted the abundancies of different isotopic CO2 species that would result from the previously proposed LH mechanisms 2,19   the respective mechanisms (Scheme 1 -Scheme 3).We note that for the previously proposed LH mechanisms, there could be scrambling reactions (of 16 O and 18 O) between different reaction intermediates (such as adsorbed O, CO, O2, OOH, OH, and H2O).Combinations of such scrambling reactions could possibly take place so that the predicted isotopic abundance resulting from the respective LH mechanisms would be close to the experimentally observed abundances of 50 % C 16 O2 and 50 % C 16 O 18 O.We have carried out additional DFT calculations on some potential 18 O exchange pathways that could change the predicted isotopic CO2 abundances in the LH mechanisms (Table S3).
However, none of these potential pathways appear plausible compared to the much lower reaction barriers in the respective LH mechanisms.We therefore believe the predictions (Scheme S1 -Scheme S3) of the isotopic CO2 abundances for the mechanisms evaluated in this manuscript are robust.
Moreover, since the predicted abundance of 50 % C  16 O2 and H2 18 O.Where more than one isotope can react in an elementary step, this has been indicated by the appropriate fraction of the stoichiometric coefficient.After summing up the overall catalytic cycle, the overall reaction has been multiplied with a factor of 2, so that all stochiometric coefficients are whole numbers.* denotes an active site on the Au NP.Also consider Figure S14 and Figure S15, for further illustration of isotopic CO2 abundancies that are expected to result from Iglesia's proposed mechanism.
Figure S1.As a consequence of step (ii), the obtained structure keeps the periodicity of Au(111) along a direction, while the periodicity is removed along b direction.A similar procedure has been appliedfor MoO2/Pt composites.15

Figure S4 .
Figure S4.Elementary reactions used for DFT calculation of reaction energetics for Iglesia's proposed LH mechanism 2 (Scheme 2, main paper and TableS2), occurring on Au(111).Note that some reactions (R12, R13, R14) were divided into several elementary steps, to enable DFT computation. .Reaction free energies and activation free energies in eV.Structures labelled 1. -11.

Figure S5 .Figure S6 .
Figure S5.Elementary reactions used for DFT calculation of the reaction energetics (summarized in Table S2) for our proposed w-MvK mechanism of water-promoted CO oxidation over Au-γ-Fe2O3.The mechanism is also discussed in detail in the main paper, (Figure 3 and Scheme 3) and is represented by R15, R17 -R21.Direct abstraction of lattice oxygen by CO (non-water promoted) is represented by R16.Reaction free energies and activation free energies in eV.Note that some reactions (R17, R19, R21) were divided into several elementary steps, to enable DFT computation.Reaction free energies and activation free energies in eV.Structures labelled 1. -12.

Figure
Figure S7.(A) Base-line corrected calibration curve for CO2 concentration in our reactor, as measured by mass spectrometry.(B) Typical mass-spectrum during measurement of CO oxidation rate over Au-γ-Fe2O3 in absence of O2 (see Figure 2 B for rate data).The concentration of CO2 measured in this spectrum is (using the calibration curve in (A)) 190 +/-2 ppm (95 % confidence).

Figure
Figure S8.(A) Base-line corrected calibration curve for H2 concentration in our reactor, as measured by mass spectrometry.(B) Transient CO2 (blue circles) and H2 (green squares) production rates over Au-γ-Fe2O3 during CO oxidation in absence of O2.Reaction conditions: 1 vol % CO, 2.8 vol % H2O, balance N2.Reaction temperature was 25 o C, and pressure 1 atm.The gas hourly space velocity (GHSV) was 21 L h -1 g -1cat, and the CO conversion was below 20 % (ensuring data was collected under differential conditions).2,3Note that the CO2 rate data is the

Figure
Figure S9.(A) Full data-sets for Figure 2 A, main paper, describing transient CO oxidation rates over Au-TiO2.(B) Full data-sets for Figure 2 B, main paper, describing transient CO oxidation rates over Au-γ-Fe2O3.Refer to Figure 2 for reaction conditions.Refer to Figure S7 A for the CO2 calibration used to determine CO2 concentrations in the reactor effluent.Note that for Au-TiO2, no significant CO2 production was observed in absence of O2, and therefore this measurement was only repeated two times.

Figure S11 .
Figure S11.Examples of HAADF-STEM micrographs of the Au-γ-Fe2O3 catalyst.After investigation of many (>20) such micrographs, we could not find any evidence of single Au atoms, or very small Au-clusters.We therefore conclude that Au NPs in the range 2 -8 nm (Figure 1 D) are the dominating Au-species in the catalyst.
O2 molecule transforms either to C 16 O 18 O or C 18 O2 by single (Equation S6) or double 18 O (Equation S7) exchange with H2 18 O: Pexch.(C 16O2 ⟶ C 16 O 18 O) = 0.359 (S6) Pexch.(C 16O2 ⟶ C 18 O2) = 0.066 (S7) O exchanges (Equation S8), Pexch.(C 16O 18 O ⟶ C 18 O2) = 0.359 / 2 + 0.066 = 0.246 (S8) Now, let us predict the relative abundances of the CO2 isotopic species resulting from C 16 O oxidation with 16 O2 and H2 18 O assuming the different reaction mechanisms considered in this paper.First, we consider the LH mechanism proposed by Chandler et.al. (Scheme 1 in main paper).Here we adapt this scheme to describe the reaction involving C 16 O, 16 O2 and H2 18 O (Scheme S1).In this mechanism, water only participates via a series of proton exchanges, and no 18 O incorporation into CO2 is expected.As a consequence, Chandler's LH mechanism would result in 100 % C 16 O2.Next, consider the LH mechanism proposed by Iglesia et al. (scheme 2 in main paper).Here we adapt this scheme to describe the reaction involving C 16 O, 16 O2 and H2 18 O (Scheme S2).By studying scheme S2 we conclude that Iglesias's LH mechanism would result in 75 % C 16 O2 and 25 % C 16 O 18 O.Finally, in Scheme S3, consider the w-MvK mechanism proposed in this work (Scheme 3 in main paper) adapted to the reaction involving C 16 O, 16 O2 and H2 18 O.By analyzing Scheme S3, we conclude that our proposed w-MvK mechanism would result in 50 % C 16 O2 and 50 % C 16 O 18 O.To further illustrate the CO2 isotopic abundancies that should be expected from the different mechanisms, consider Figure S13 -Figure S16,

Figure S12 .
Figure S12.Time evolution of the relative fractions (%) of C 16 O2, C 16 O 18 O and C 18 O2 in the C 16 O2 + H2 18 O control-experiment to determine reactorspecific probability of transformation of different CO2species into other CO2species by 18 O exchange with H2 18 O.

Figure S13 .
Figure S13.Reproduction of Figure S2, but with H2 18 O instead of H2 16 O.The figure illustrates how the abundancies of different isotopic CO2 species can be predicted from Chandler's proposed reaction mechanism 19 (Scheme 1 main paper, and Scheme S1) of water-promoted CO oxidation with C 16 O, 16 O2 and H2 18 O.It can thus be predicted the mechanism would result in 100 % C 16 O2.Reaction free energies and activation free energies in eV.Structures labelled 1. -13.

Figure S14 .
Figure S14.Reproduction of Figure S3, but with H2 18 O instead of H2 16 O.The figure illustrates how the abundancies of different isotopic CO2 species can be predicted from Iglesia's proposed reaction mechanism 2 (Scheme 2 main paper, and Scheme S2) of water-promoted CO oxidation with C 16 O, 16 O2 and H2 18 O.Here illustrated allowing the mechanism to run on both the Au NP and the γ-Fe2O3 support.It can thus be predicted the mechanism would result in 75 % C 16 O2 and 25 % C O 18 O.Reaction free energies and activation free energies in eV.Structures labelled 1. -11.

Figure S15 .
Figure S15.Reproduction of FigureS4, but with H218 O instead of H216 O.The figure illustrates how the abundancies of different isotopic CO2 species can be predicted from Iglesia's proposed reaction mechanism 2 (Scheme 2 main paper, and Scheme S2) of water-promoted CO oxidation with C 16 O,16 O2 and H218 O.Here with the mechanism running on Au(111).It can thus be predicted the mechanism would result in 75 % C 16 O2 and 25 % C 16 O 18 O (as is the case if the mechanism is allowed to run on Au-γ-Fe2O3, see FigureS14).Reaction free energies and activation free energies in eV.Structures labelled 1. -11.

Figure S16 .
Figure S16.Reproduction of Figure S5, but with H2 18 O instead of H2 16 O.The figure illustrates how the abundancies of different isotopic CO2 species can be predicted from our proposed w-MvK reaction mechanism 2 (Scheme 3 and Figure 3 main paper, and Scheme S3) of water-promoted CO oxidation with C 16 O, 16 O2 and H2 18 O.It can thus be predicted the w-MvK mechanism would result in 50 % C 16 O2 and 50 % C 16 O 18 O.Reaction free energies and activation free energies in eV.Structures labelled 1. -11.

Table S1 :
Mismatch parameter values (ξ) for different γ-Fe2O3 surfaces with Au(111).ξ Values were calculated as indicated in Equation S1. a, b, c, α, β, and γ are the surface structure parameters obtained using ASE 6 python package.  and   are the surface area, and  / is the overlap surface area.
2  +     2 +     Here the values {  ,   , and   } are the number of reacting molecules and the values {  , and   } are the number of product molecules.Note that water is in both the reactant and product, since water is not present in the global reaction, "2CO(g) + O2(g) → 2 CO2 (g)".Thermal corrections to free energies were obtained from the frequency analysis by assuming harmonic vibrations.Vibrational contributions were included for the surface states allowing only the adsorbates to vibrate. +/− 2  3 and  /− 2  3 , Gibbs free energies were approximated by the sum of the internal DFT/PBE energy () and zero-point energy (), e.g.,  /− 2  3 =  /− 2  3 +  /− 2  3 .

Table S2 .
29mmary of DFT calculated reaction energetics for our proposed w-MvK mechanism and for previously proposed LH mechanisms by Chandler19and Iglesia.2Everyreactionelementarystep is shown in FiguresS2-S5.
Figure S2.Elementary reactions used for DFT calculation of reaction energetics for Chandler's proposed LH mechanism We wish to point out that in predicting the abundances expected to result from the different mechanisms, we assume that no other elementary reactions occur other than what is explicitly listed in 16O2 50 % C16O18O emerges from the w-MvK mechanism without making any additional assumptions about scrambling reactions, we believe the w-MvK mechanism offers the most straightforward rationalization of our experimental data.Elementary reaction steps for the (by Chandler et al.) 2 postulated LH-reaction mechanism of waterpromoted CO oxidation over Au-Al2O3, Au-TiO2, and later postulated over Au-Fe2O3.20.Here we adapt Scheme 1 (main paper) to consider the reaction with C 16 O,16O2 and H218O.*denotes an active site on the Au NP, away from the NP-support interface, † denotes an Au site at the NP-support interface, and ‡ denotes a support site at the NP-support interface.Also nsider FigureS13, for further illustration of isotopic CO2 abundancies that are expected to result from Chandler's proposed mechanism.Elmentary reaction steps for the LH-reaction mechanism postulated by Iglesia et al. of waterpromoted CO oxidation over Au-Al2O3, Au-TiO2, and Au-Fe2O3.2HereweadaptScheme 2 (main paper) to consider the reaction with C16O,